Plano, Texas · Senior Business Analyst · Salesforce · AWS · AI

Turning enterprise requirements into systems that ship, integrate cleanly, and prove their value.

Senior Business Analyst at HotelKey, working at the scope of a solution architect. Five-plus years across hospitality SaaS, automotive, and financial services, sitting where enterprise clients, product, and engineering meet: running discovery, designing the Salesforce, AWS, and API integrations that follow, testing them to certification, and increasingly wiring GenAI into the delivery itself.

4,000+Properties on the HotelKey platform
9+Enterprise brands · Hilton, IHG
30%Fewer API support tickets
95%Defect-free deployments
01About

Analyst by title, architect by scope.

Most of my work happens in the gap between what a hotel brand's operations team needs and what an engineering team can ship. I run the discovery sessions, write the user stories and data-mapping rules, build the Postman collections that prove a CRS integration works, and walk clients through UAT and partner certification until the feature matches the logic we agreed on. Then I do it again for the next brand. The title says analyst; the day-to-day is the solution design, the integration decisions, and owning that they hold up in production.

On the platform side I'm a Salesforce specialist (Apex, Lightning Web Components, Flows, Einstein) with five certifications and a preference for declarative solutions that survive the next admin. On the cloud side I work in AWS daily: reservation data products on Kinesis, Glue, and Athena, and API performance against DynamoDB-backed services. On the data side I write the SQL and Athena queries that audit what the APIs actually did.

Every engagement I've led carries a number next to its name. If it can't be measured, I haven't finished.

Lately the interesting question is what agentic AI looks like inside a governed enterprise platform. My answer-in-progress is Trowel, an AI agent that reads a Salesforce org's metadata and produces the kind of Well-Architected assessment consultancies bill five figures for. The other eleven projects below are the same instinct pointed at APIs, delivery, and hospitality data.

Currently

Week 2 of an 8-week build of Trowel for Salesforce · studying for Salesforce AI Specialist · shipping one project a week through September.

02Experience
2023 to NowAug 2023 – Present

Senior Business Analyst

HotelKey Inc. · Carrollton, TX · promoted from Business Analyst · solution-architecture scope across CRS, Salesforce, and AWS
  • Primary analytical liaison for 9+ enterprise hotel brands (Hilton, IHG) on a platform serving 4,000+ properties, covering discovery, data mapping, user stories, and functional specs. Implementation effort down 30% post-go-live.
  • Architected a Salesforce Sales & Support platform on Einstein AI and automated Flows, delivering +20% lead conversion with First Contact Resolution held above 70%.
  • Integration testing and release-readiness across 10+ mission-critical CRS APIs with Postman collections and Swagger/OpenAPI specs, resulting in 30% fewer API support tickets.
  • Supported delivery of enterprise reservation data products on AWS (Kinesis, Glue, Athena), validating near-real-time booking data with Athena and SQL; contributed to performance validation of a DynamoDB Accelerator (DAX) rollout against enterprise SLAs.
  • SQL audits of reservation and booking data to resolve client-reported discrepancies; UAT and partner certification cycles against CRS sandbox environments.
  • Tracked integration SLAs (uptime, latency, error rates) and drove root-cause resolution; GenAI reporting cut program review cycles 20%.
2023Jan – May 2023

Financial Business Analyst

CG Infinity · Plano, TX
  • GenAI-augmented market insights with Python, Alteryx, and LLM summarization, driving +8% profit and two strategic initiatives launched.
  • AI-assisted forecasting in Power BI on Snowflake, cutting manual analysis time 35% and surfacing executive insights 2× faster.
  • Snowflake pipelines with automated data-quality checks for +30% query performance; predictive risk models cut exposure 10%.
2019 to 2021Sep 2019 – Dec 2021

Systems Engineer · Salesforce Business Analyst

Infosys Limited · Bangalore, India · client: Daimler AG
  • Led Salesforce re-architecture for Daimler AG (Mercedes-Benz) with Apex and LWC, reducing manual entry 20% and improving data accuracy 15%.
  • Established CI/CD with GitHub Actions, Bitbucket, and Jenkins, achieving 95% defect-free deployments with post-release issues down 30%.
  • Led full-cycle UAT, onboarding, and hypercare to 90% CSAT with resolution time cut 40%; executive dashboards tied to a 23% revenue increase.
03Case studies

Three pieces of work, in detail.

04Projects

Twelve things I'm building.

A public build log, not a trophy case. Each entry ships to GitHub when it meets its own definition of done; repos flip from coming soon to live as they land. Click a row for scope, stack, and the impact it's designed to have.

01 Trowel for SalesforceAI agent that audits a Salesforce org and writes the Well-Architected assessment AI · Salesforce Python · Claude API · sf CLI · networkx · SQLite In progress
What it does

Excavates an org's metadata (Flows, Apex, fields, permission sets) into a dependency graph, then an agent with hybrid retrieval (graph traversal + embeddings) finds dead automation, conflicting triggers, and permission sprawl and narrates a prioritized remediation roadmap. Findings are detected deterministically; the LLM explains, never invents.

Impact

Replaces a $50k+ manual org-health assessment with a reproducible report in minutes, gated by a golden-dataset eval suite in CI.

Skills
Agentic AIRAG / hybrid retrievalSalesforce Metadata APIGraph analysisLLM evalsWell-Architected
GitHub PakshalShah30/trowel-for-salesforce · README up, code landing weekly
02 FlowLintA static analyzer for Salesforce Flow XML, like ESLint for Flows Salesforce · API Python · lxml · CLI · GitHub Action Coming soon
What it does

Parses .flow-meta.xml and flags the things that hurt in production: missing fault paths, hardcoded record IDs, DML inside loops, recursion risk from record-triggered Flows updating their own object, and inactive-but-referenced Flows. Runs locally or as a CI check on every pull request.

Impact

Catches the top five Flow defects before deployment instead of after a support ticket, the same class of issue that drove 30% of API support volume at scale.

Skills
Flow metadataStatic analysisCI/CDPythonDeveloper tooling
GitHub PakshalShah30/flowlint · coming soon
03 CRS Certification HarnessAutomated partner-certification test cycles for hospitality reservation APIs API · Product Postman · Newman · Python · HTML reports Coming soon
What it does

Codifies a CRS partner-certification cycle as a versioned Postman collection with data-driven test cases, runs it against sandbox/staging via Newman, and generates the pass/fail certification document reviewers actually want, with payload diffs for every failure.

Impact

Turns a multi-day manual certification into a repeatable one-command run with an audit-ready report.

Skills
Integration testingPostman / NewmanREST APIsCRS / hospitality distributionTest documentation
GitHub PakshalShah30/crs-certification-harness · coming soon
04 SLA SentinelIntegration uptime, latency, and error-rate monitor with contractual-SLA breach alerts API · Product Python · FastAPI · SQLite · Chart.js Coming soon
What it does

Probes a set of REST endpoints on a schedule, stores response time and status, computes rolling uptime and p95 latency against per-integration SLA thresholds, and posts breach alerts to Slack or email. A small dashboard shows the last 30 days per partner.

Impact

Makes SLA tracking a system instead of a spreadsheet, so escalations happen on data, not on the client noticing first.

Skills
SLA managementObservabilityFastAPIAlertingDashboards
GitHub PakshalShah30/sla-sentinel · coming soon
05 Story-to-FlowUser story + acceptance criteria in, deployable Flow metadata and Gherkin tests out AI · Salesforce Claude API · Python · Flow metadata · Gherkin Coming soon
What it does

A BA writes a story in plain English; the tool generates a record-triggered Flow definition, a Gherkin acceptance-test file, and a human-review diff before anything is deployed. Schema-forced output keeps the metadata valid; a review step keeps the human in the loop.

Impact

Compresses the story-to-working-automation loop from days to an afternoon while producing the test artifacts UAT needs anyway.

Skills
Requirements engineeringPrompt engineeringStructured outputFlow BuilderBDD / Gherkin
GitHub PakshalShah30/story-to-flow · coming soon
06 RAID RadarMeeting notes and Jira in, a living Risks / Actions / Issues / Decisions log out AI · Product Claude API · Python · Jira REST · Markdown Coming soon
What it does

Ingests meeting transcripts and ticket updates, extracts RAID items with owners and dates, de-duplicates against the existing log, and flags risks that resurface across sprints. Produces the weekly status draft a program manager would otherwise write by hand.

Impact

Designed to cut program-review prep time by a further 20% on top of what GenAI reporting already delivered at HotelKey.

Skills
Program managementInformation extractionJira APIAgentic workflows
GitHub PakshalShah30/raid-radar · coming soon
07 Property PulseCross-property support-ticket clustering and root-cause drafts for multi-brand hotel platforms AI · Product Python · sentence-transformers · scikit-learn · Claude API Coming soon
What it does

Embeds and clusters support tickets across properties and brands, separates "one property's configuration error" from "platform-wide defect," and drafts the RCA document with the affected-property list attached. Built on synthetic hospitality ticket data.

Impact

Surfaces platform defects days earlier by seeing the pattern across 1,000 properties that no single support agent can.

Skills
EmbeddingsClusteringRoot-cause analysisHospitality operations
GitHub PakshalShah30/property-pulse · coming soon
08 Prompt Lab for EinsteinAn evaluation harness for Salesforce Prompt Builder templates AI · Salesforce Python · Prompt Builder · LLM-as-judge · pytest Coming soon
What it does

Versions Prompt Builder templates alongside a golden dataset of inputs and expected outputs, scores each template version with a rubric-driven judge model, and reports regressions before a template is promoted. The eval discipline from Trowel, pointed at Einstein.

Impact

Gives admins a way to change a production prompt with evidence instead of vibes, which is the missing piece in most Agentforce rollouts.

Skills
Einstein / AgentforceLLM evaluationPrompt engineeringRegression testing
GitHub PakshalShah30/einstein-prompt-lab · coming soon
09 Data MapperSchema-to-schema mapping specs with auto-generated SQL validation queries API · Product Python · SQL · JSON Schema · Markdown Coming soon
What it does

Takes two system schemas (say, a CRS reservation payload and a CRM object) and a mapping definition, then emits the data-mapping document for stakeholders and the SQL queries that verify the mapped data landed correctly. One source of truth for the BA doc and the audit.

Impact

Ends the drift between the mapping spec everyone signed and the data that actually arrives.

Skills
Data mappingSQL auditingIntegration designDocumentation automation
GitHub PakshalShah30/data-mapper · coming soon
10 Lead Router APIA rules-driven lead assignment service that syncs to Salesforce via REST and webhooks API · Salesforce FastAPI · OAuth 2.0 · Salesforce REST API · PostgreSQL Coming soon
What it does

Accepts leads from any source, applies configurable routing rules (territory, capacity, round-robin, score thresholds), writes assignments to Salesforce through the REST API with OAuth, and exposes a webhook so downstream systems learn about the assignment in real time. Postman collection and OpenAPI spec included.

Impact

Demonstrates the full API product lifecycle (design, auth, docs, tests, versioning) on a workflow every sales org has.

Skills
API designOAuth 2.0Salesforce REST APIOpenAPIWebhooks
GitHub PakshalShah30/lead-router-api · coming soon
11 Forecast KitA reproducible Snowflake + dbt + Power BI forecasting pipeline template Product · Data Snowflake · dbt · Python · Power BI Coming soon
What it does

Packages the forecasting pattern from my CG Infinity work as a template: dbt models with data-quality tests, a Python forecasting layer, and a Power BI report wired to Snowflake, with a seed dataset so anyone can run it end to end.

Impact

The version of "35% less manual analysis time" that someone else can clone and stand up in an afternoon.

Skills
SnowflakedbtPower BIForecastingData quality
GitHub PakshalShah30/forecast-kit · coming soon
12 Onboarding Playbook GeneratorBrand requirement docs in, property-onboarding checklists and UAT scripts out AI · Product Claude API · Python · DOCX generation Coming soon
What it does

Reads a hotel brand's integration requirements and produces the onboarding playbook: property-level checklist, configuration steps, UAT scripts with acceptance criteria, and a go-live sign-off sheet, delivered as a formatted Word document the client can actually use.

Impact

Standardizes the onboarding artifacts behind the 30% implementation-effort reduction so every new brand starts from the same playbook.

Skills
Client onboardingUAT designDocument generationHospitality SaaS
GitHub PakshalShah30/onboarding-playbook · coming soon
05Capabilities

What I bring to a room.

Salesforce platform5× certified

Apex, Lightning Web Components, Flows, Einstein AI, Data Cloud. Well-Architected by habit: declarative first, code where it earns its place, CI/CD around all of it.

AWS & cloud dataCertified Cloud Practitioner

Kinesis, Glue, Athena, DynamoDB, S3, Lambda, API Gateway. Reservation data pipelines and near-real-time booking visibility; validating API performance and SLAs on DynamoDB-backed services.

Integration & APIsREST · SOAP · OAuth

Postman collections, certification cycles, SLA monitoring, payload auditing with SQL. I treat an integration as untested until the client's data has landed and been queried.

Applied AILLM · RAG · agents

Claude, Einstein, prompt engineering, hybrid retrieval, LLM-as-judge evals. The interesting part isn't the model; it's the eval gate that makes it safe to ship.

Analysis & deliverySenior BA · solution design

Discovery, user stories, acceptance criteria, data mapping, UAT, partner certification, backlog prioritization, Agile/Scrum, Jira & Confluence. Client-facing by default; every requirement traceable to a test and a metric.

DataSQL · Snowflake · BI

SQL, Snowflake, dbt, Power BI, Tableau, Python, Alteryx. Pipelines with quality checks, dashboards leadership actually opens.

06Credentials

Verified, not self-declared.

SalesforceCertified Platform Developer II
SalesforceCertified Data Cloud Consultant
SalesforceCertified Business Analyst
SalesforceCertified Administrator
SalesforceCertified Platform Developer I
Amazon Web ServicesCertified Cloud Practitioner
2022 – 2023M.S. Business Analytics, The University of Texas at Dallas
2016 – 2019B.E. Electronics Engineering, University of Mumbai

Verify: Salesforce Trailblazer profile · AWS on Credly

07Writing

Notes from the build.

Drafting

Why pure vector RAG gives wrong answers on Salesforce metadataStructured data needs graph traversal for facts and embeddings for semantics. What broke when I tried RAG alone, and the hybrid retriever that fixed it.

In draft
Drafting

What I learned reading Salesforce Flow XML by handStatus, processType, start blocks, recordUpdates: the six elements a static analyzer actually needs, and the dead-Flow definition that follows from them.

In draft
Planned

"It works on my side" is not a statusHow written pass/fail certification and SQL audits cut API support tickets by 30%, and the template I'd hand any integration analyst.

Planned
08Contact

If you need someone who writes the spec and can defend the architecture.

I read everything that comes in. Fastest replies go to specific questions: a role, a project, an org that needs an honest look.